Analysis of Common Bubbles in Financial MarketsUsing Mixed Causal-Noncausal VAR Modelswith Heavy-Tailed Distributions

Abstract

This article explores the phenomenon of common financial bubbles in heterogeneous markets—such as equities, cryptocurrencies, and commodities—using advanced mixed causal-noncausal Vector Autoregression (VAR) models with heavy-tailed distributions. This approach captures not only traditional causal relationships between assets, but also complex and explosive dependencies that often emerge in contexts of speculative exuberance. The research is based on evidence that conventional linear models often fail to identify bubble episodes, primarily due to the frequent presence of extreme events in financial series. Therefore, Student's t-distributions and regime-switching methodologies are employed, which facilitate a robust analysis of local dynamics and the simultaneity of bubbles across assets. The results demonstrate that bubbles can emerge in a coordinated manner, revealing patterns of contagion and interdependence that are critical for risk management and financial policymaking. The study concludes that mixed VAR models, by integrating causal and non-causal elements under the assumption of heavy tails, represent a significant improvement for both the theoretical understanding and practical monitoring of financial stability.